Architecture

End-to-end call flow

sequenceDiagram
  participant User as SIP handset
  participant PBX as Your PBX
  participant Ex as sip-dmr-exchange
  participant Br as sip-dmr-bridge
  participant DMR as DMR / SystemX server

  User->>PBX: Dial feature code + enter TG
  PBX->>Ex: INVITE PJSIP/TG@sip-dmr-exchange + headers
  Ex->>Br: Docker spawn + SIP B2BUA
  Br->>DMR: MMDVM login
  Ex-->>PBX: 200 OK after bridge healthy
  User<<->>DMR: Audio via exchange

Components

Layer	Typical host	Responsibility
IVR + access	PBX	Feature code, talkgroup entry, allowlist
SIP trunk	PBX	PJSIP/${TG}@sip-dmr-exchange
Exchange	Portal VM	Auth, provision bridge, B2BUA media
Bridge	Portal VM (Docker)	One container per call; MMDVM client
Dashboard	Portal VM (optional)	Static UI + nginx proxy to HTTP API

SIP headers (PBX → exchange)

Header	Purpose
X-Auth-Token	Shared secret; exchange returns 403 if missing/wrong
X-Callsign	Caller callsign for bridge naming and RadioId lookup
X-DMR-Id	Optional explicit DMR ID
X-DMR-SSID	Optional SSID suffix for SystemX client ID

Talkgroup is the SIP request URI user part (the ${TG} in PJSIP/${TG}@trunk).

Exchange internals

• Listens on SIP_BIND_ADDR:SIP_BIND_PORT (default 0.0.0.0:5060)
• On INVITE: provisions sip-dmr-bridge:latest via local Docker socket
• Holds call until bridge HEALTHCHECK passes (~5–15 s cold start)
• Bridges RTP between incoming PBX leg and bridge leg (B2BUA)
• Enforces MAX_ACTIVE_CALLS — returns SIP 503 when at capacity

HTTP API

Optional nginx (or similar) can expose:

GET /bridges

Active bridge sessions (no auth on read in current implementation):

[{
  "callsign": "M0ABC",
  "dmr_id": 1234567,
  "talkgroup": 234,
  "state": "Active",
  "state_time": "2026-08-16T12:00:00",
  "active": true
}]

Bridge states: Inactive, Provisioning, Provisioned, Active, Deprovisioning, Deprovisioned.

GET /stats

{
  "uptime": 12345.6,
  "cpu_percent": 12.3,
  "memory_percent": 75.0,
  "calls": 42,
  "users": 7
}

Per-call resource cost

Measured on a small portal VM:

• ~130–150 MB RAM per bridge container (includes md380-emu)
• ~10–15% CPU per active bridge under TX/RX
• ~5–15 s until bridge healthy (caller may hear ringing)

Size your VM accordingly; see Scaling.

Related

• Self-hosting overview
• Configuration
• FreePBX integration

---

Last update: August 17, 2026 21:08:54
Last updated by: MØVUB

---

Help improve this page by contributing to our documentation.